Uncertainty is a key part of any sports game; without it, there is little reason to be interested in the outcome. This thesis attempts to quantify the uncertainty inherent in NHL hockey games by building a real-time win probability model that estimates both teams’ likelihood of winning based on what has happened in the game so far. The model is built using historical data from the 2009-2010 season all the way to the 2016-2017 season. Given the differential and the time left, the model evaluates historical data for that specific game-state and calculates a win probability. The model also uses a multi-regression approach to incorporate pre-game Vegas odds as a way to factor the strength of both teams; to my knowledge, this is the first publicly available hockey win probability model to do so. Finally, the model also factors in elements unique to the sport of hockey, like power plays and shootout periods.